The teaching persona controls how your Bloom talks to students: how much it questions versus tells, how much writing it generates, whether it confirms answers, and how much code it produces. You set this on the Pedagogy tab inside a Bloom’s configuration. Choose the Default preset or switch to Custom to adjust five independent dimensions (Writing Feedback, Writing Generation, Solution Process, Answer Provision, and Code Assistance).
Choose support based on prior knowledge, task complexity and your learning goals. Novices may need explicit explanations and worked examples before attempting a task; learners with more experience may benefit from less guidance and more independent practice. Questioning and hints are useful options, not inherently better teaching. See our learning science principles for the research behind this approach.
Persona changes apply when a student sends their next message, including in an existing conversation. Earlier replies stay as they were. These controls apply to student conversations; use View as Student to preview them as an educator.
Before you start
- A Bloom you have already created. If not, see Create your first Bloom.
- A manager role on that Bloom, or Space Owner access, and a paid teaching-space plan. Students and users with only the educator role cannot change pedagogical settings. If the controls are locked, check the explanation shown or ask your Space Owner for help.
Step 1: Open the Pedagogy tab
From the Manage Blooms dashboard, click the Bloom you want to edit, then click Configuration in its sidebar, then choose Pedagogy from the configuration tabs. On wider screens, these tabs appear down the left side; on smaller screens, they appear in a scrollable row above the settings.
Step 2: Choose Default or Custom
You will see two preset cards:
- Default. “Standard learning support settings. Switch to Custom to adjust the guidance available to students.” This is the current product preset, not a recommendation for every learner or task. Its settings include light grammar suggestions, structural scaffolds (no prose), setup-only guidance for problems, a correct/incorrect signal without revealing the answer, and partial code snippets.
- Custom. “Fine-grained control over how much help the AI provides. Adjust each dimension independently.” Switching to Custom for the first time pre-populates every dimension with the calibrated defaults, so you only need to change the ones you care about.
Step 3: Tune the five dimensions
Each dimension has four levels, from most restrictive (left) to most permissive (right). The segmented buttons show the level name; the text below the control describes the guidance allowed at that level. More restrictive does not necessarily mean more effective for learning.
Writing Feedback
How much the AI can modify the student’s writing.
- Comments only. High-level observations about structure, evidence, and tone. No specific word or phrase suggestions.
- Rephrase. Grammar fixes and wording tweaks at sentence level, framed as suggestions. (Default)
- Sentence rewrite. May suggest clearer phrasing for individual sentences. One rewrite per response.
- Paragraph rewrite. May rewrite full paragraphs in the student’s voice, using only their ideas.
Writing Generation
How much new prose the AI can produce.
- None. No new prose. Only feedback on the student’s existing work.
- Scaffold. Bullet outlines, structural plans, topic and evidence slot templates. No prose. (Default)
- Example sentences. Model phrases and sentences to learn from. Labelled as examples.
- Example paragraphs. Complete model paragraphs as exemplars for learning.
Solution Process
How much working and steps the AI reveals.
- Conceptual hints. Name the relevant method or theorem. No calculation steps.
- Setup guidance. Help frame the problem, identify variables, choose method. Stop before solving. (Default)
- Partial working. Show the method step-by-step but stop before the final answer.
- Full solution. Complete step-by-step solution including the final answer.
Answer Provision
Whether the AI confirms or reveals answers.
- No confirmation. Ask students to check their reasoning without confirming whether their answer is correct.
- Correct / incorrect. Say “correct” or “not quite” without revealing the answer. (Default)
- Guided reveal. Confirm when correct. Hints when wrong. Reveal after multiple attempts.
- Direct answer. Provide the correct answer when asked or after a failed attempt.
Code Assistance
How much code the AI can write.
- Conceptual only. Describe approach in words. No code of any kind.
- Pseudocode. Algorithmic structure, no runnable syntax.
- Partial code. Key functions or snippets. Student completes the rest. (Default)
- Full implementation. Complete working code with explanation.
Common issues
The dimension controls are greyed out
You are still on the Default preset. Click the Custom card to enable them. If Custom is also disabled, check the message shown: you need a paid plan and a role that allows configuring this Bloom.
An existing conversation still shows the previous teaching style
Changes apply when the student sends their next message. If a student has not sent a message since the change, they will still see the prior style on screen. The new rules apply on their next reply.
Educators are not seeing any persona constraints
These settings shape the student experience. Educators can still get direct answers when preparing lessons. Use View as Student in the top-right of the chat to test the persona from a student’s perspective.
What’s next
- Lock the AI to specific topics → Controlling topics
- Build a practice activity from your materials → Create quizzes
- Open Bloom and tune your persona now → Manage Blooms